LncLocation: Efficient Subcellular Location Prediction of Long Non-Coding RNA-Based Multi-Source Heterogeneous...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• The subcellular localization of long non-coding RNAs (lncRNAs) is critical for elucidating their biological functions.
• Limited experimentally verified data and imbalanced distribution across organelles present significant challenges for predicting lncRNA localization.
• Existing prediction methods struggle with the multi-classification, small-sample imbalance problem inherent in lncRNA subcellular localization prediction.

Purpose of the Study:

• To develop a robust computational tool, lncLocation, for predicting the subcellular location of lncRNAs.
• To address the challenges posed by small sample sizes and data imbalance in lncRNA localization prediction.
• To improve the accuracy and reliability of lncRNA subcellular localization prediction using integrated multi-source features.

Main Methods:

• Integration of multi-source features to construct a sequence-based computational tool.
• Utilization of Autoencoder for feature enhancement.
• Application of binomial distribution-based filtering and recursive feature elimination (RFE) for feature selection.
• Comprehensive experimentation with feature combinations and machine learning models to select optimal parameters.

Main Results:

• The developed lncLocation tool achieved an accuracy of 87.78% using 5-fold cross-validation on benchmark data.
• lncLocation outperforms existing state-of-the-art tools in predicting lncRNA subcellular localization.
• Significant improvements in classification performance were observed, particularly for underrepresented lncRNA classes.

Conclusions:

• lncLocation provides an effective solution for predicting lncRNA subcellular localization, overcoming data imbalance issues.
• The integrated feature approach and optimized machine learning model enhance predictive accuracy.
• This tool advances the field by offering a more reliable method for understanding lncRNA function through localization prediction.